🏆
Global Research Authority
Serving Researchers Since 2012

Inventory Optimization and Just-in-Time Material Planning for Cost Reduction in EV Battery Pack Manufacturing: An Applied Study at Endurance Technologies Ltd., New Energy Plant

DOI : 10.17577/IJERTCONV14IS120008
Download Full-Text PDF Cite this Publication

Text Only Version

Inventory Optimization and Just-in-Time Material Planning for Cost Reduction in EV Battery Pack Manufacturing: An Applied Study at Endurance Technologies Ltd., New Energy Plant

Ayusha Vaibhav Homkar

Student, MBA- Operations and Supply Chain Management, School of Business,

Indira University, Pune, India 41103

Dr. Nagesh Tarte

Associate Professor – MBA, School of Business, Indira University,

Pune, India 411033

Mr. Mohit Sinha,

Plant Head, Endurance Technologies Ltd., New Energy Plant,

Pune, Maharashtra, India.

Dr. Manisha Maddel

Associate Professor & Guide, School of Business, Indira University,

Pune, India-411033 ORCID: 0009-0008-1720-2959

Abstract – Electric-vehicle (EV) battery packs are costly, safety-critical products. Their manufacturers must keep assembly lines supplied without tying up working capital or floor space and without exposing stock to expiry, obsolescence, and customer rejection. This applied study, completed during on- the-job training at the Endurance Technologies Ltd. New Energy Plant in Pune, builds an integrated inventory and just-in-time (JIT) planning framework that connects demand, capacity, bill-of- materials (BOM) requirements, supplier ordering, production planning and control (PPC), and special- inventory handling. The evidence comprises a 10,000-pack monthly baseline, a four-hour line-side JIT model, illustrative cost scenarios, an import lead-time map, PPC records for 12 dates in July August 2026, a 25 August 2026 stores snapshot, and plant counts of 400 obsolete and 250 customer- rejected packs. The baseline uses 55.56% of rated capacity, and stacked buffers inflate an 800,000-unit requirement to 906,576 units (+13.3%). PPC achievement was 50.82% in July and 69.64% in August on the recorded dates, and imported material carries a 10-day gap between 55-day lead time and 45-day coverage. The paper introduces an expiry- safe achievement threshold (A* = S ÷ [D × u]) and a JIT break-even carrying rate (27.2% per year in the base case). Simulation-calibrated regressions, reported as model-based evidence, show that plan adherence outweighs buffer size in explaining overstock and that blanket JIT is cost-negative in most scenarios. The study therefore recommends

controlled JIT supported by first-expired-first-out (FEFO) issue, daily PPCStores feedback, engineering-change gates, and root-cause control of customer rejections.

Keywords: inventory optimization, just-in-time, bill of materials, FEFO, production planning and control, EV battery pack

  1. INTRODUCTION

    Indias electric-vehicle market is expanding quickly. The India Brand Equity Foundation (IBEF, 2026) puts EV sales at roughly 2.66 million units in FY26 and notes continuing investment in batteries, components, and charging networks, while the PM E-DRIVE scheme, with an outlay of 10,900 crore, further encourages domestic EV manufacturing (Ministry of Heavy Industries [MHI], n.d.). Within this value chain the battery pack is among the hardest products to plan for. It is expensive, safety- critical, and fully traceable, and it is assembled from cells, electronics, enclosures, and sealants bought from many suppliers, several of them overseas. A plant therefore faces a two-sided risk. When material is missing, the line stops; when material is abundant, working capital and floor space are locked up and shelf-life-sensitive items drift toward expiry. Endurance Technologies Ltd. is an Indian automotive-component maker that has broadened its range from aluminium die casting, suspension, transmission, and braking products into lithium-ion battery packs (Endurance Technologies Ltd., n.d.). Inventory decisions at its New Energy Plant cut

    across Stores, PPC, Production, Quality, Purchase, Engineering, suppliers, and customers. Three problems surfaced repeatedly during the internship. Adhesives and sealants with limited shelf life came close to expiry whenever output fell short of plan. A batch of 400 packs built ahead of the start of production (SOP) became obsolete after the customer changed the design. And 250 packs were rejected by the customer and awaited controlled disposition. Behind all three lay the persistent gap between planned and actual output, which changes daily material consumption.

    This paper responds with an integrated, data-driven framework built in three steps. The first converts the planning baseline into capacity utilisation, BOM- level requirements, and a JIT delivery schedule, and tests their cost consequences. The second examines PPC performance, imported-material lead time, and shelf-life exposure to show where the plan and the material position diverge. The third designs separate controls for obsolete and customer-rejected inventory. Beyond plant-level results, the paper adds four contributions: formal propositions on when JIT pays and when expiry bites, a set of new decision constructs (A*, c*, LCG, BSI, and SICR), a sampling and data-quality design suited to small industrial case studies, and simulation-calibrated regression models that turn each objective into a testable relationship.

    Section 2 sets out the theory and its novelty, and Section 3 reviews the literature. Section 4 states the objectives, conceptual framework, and hypotheses. Section 5 explains the methodology, sampling, and data-quality safeguards. Section 6 analyses the plant data, Section 7 reports the regression analysis, Section 8 presents the contributions of the paper, and Section 9 concludes with findings, suggestions, and future scope.

  2. THEORETICAL BACKGROUND

    1. Core Concepts

      Just-in-time (JIT) delivers material close to the time and in the quantity it is consumed, relying on pull signals, small lots, and close supplier coordination. It does not mean zero inventory: minimum (safety) stock still guards against variation in demand, lead time, and supplier quality (Gunasekaran et al., 1993; Heizer et al., 2017). A bill of materials (BOM) turns finished-product demand into component demand through BOM explosion, and once allowances for rejection and supplier-order adjustments are added, the result is the purchase quantity. Reorder logic sets an order as lead-time demand plus minimum stock less usable stock, subject to supplier minimum order quantities (MOQ).

      For shelf-life-sensitive items, first-expired-first-out (FEFO) issues the batch with the earliest expiry first, so the value of stock depends on whether it can be consumed before it expires. PPC works as a feedback system in which actual output, and not

      only the plan, drives replenishment. Obsolescence arises when a change in design, technology, or demand leaves stock unusable, whereas customer rejection calls for a closed loop of segregation, root- cause analysis (RCA), rework and retest, and controlled disposition. Total-cost thinking weighs inventory value against transport, handling, floor space, obsolescence, and shortage risk. Finally, 5S (Sort, Set in order, Shine, Standardise, Sustain) provides the physical discipline of identification, fixed location, and cleanliness on which FEFO depends.

    2. Theoretical Lenses

      Four established theories frame the analysis. Toyota-style pull logic (Ohno, 1988; Womack & Jones, 1996) explains why small, frequent replenishment reduces waste. The theory of constraints (Goldratt & Cox, 1984) holds that system output is governed by its tightest constraint, which must be located rather than assumed. Perishable-inventory theory (Ghare & Schrader, 1963; Nahmias, 1982) treats stock as losing value with age. Littles (1961) law, wich links average inventory, throughput, and time in the system, ties days of coverage to lead time. Resilience thinking (Christopher & Peck, 2004) adds that efficiency measures such as JIT must be weighed against exposure to disruption.

    3. Theoretical Novelty

      A theoretical contribution is judged by the originality and utility it adds to an existing body of knowledge (Corley & Gioia, 2011; Whetten, 1989). The JIT, lean, and perishable-inventory streams rarely meet in one model. This paper joins them in battery-pack assembly and advances four propositions.

      Proposition 1 (conditional value of JIT). JIT lowers total cost only when the annual carrying-cost rate exceeds a break-even rate, c* = 12 × (T F)

      ÷ I, where T is the monthly increase in transport cost, F the monthly floor-space saving, and I the reduction in inventory value.

      Proposition 2 (adherence-coupled expiry). The expiry risk of a lot is a threshold function of plan adherence. A lot with stock S, planned daily use u, and D days to expiry is exposed whenever average PPC achievement A falls below A* = S ÷ (D × u). Proposition 3 (constraint migration). When utilisation of rated capacity is moderate (here 55.56%), the binding constraint migrates from capacity to material availability and plan adherence, so improvement effort should migrate with it.

      Proposition 4 (plan-blind buffer inflation). When orders follow the plan without feedback from actual output, overstock approximates (1 + BSI) ÷ A 1, where BSI = (1 + r)(1 + m)(1 + s) 1 is the stacked- buffer inflation. At BSI = 13.3% and A = 0.62, overstock is about 83%.

  3. LITERATURE REVIEW

    1. JIT and Lean Foundations

      Early reviews framed JIT as a broad philosophy. Golhar and Stamm (1991) synthesised the JIT literature and singled out supplier participation, quality management, employee involvement, and waste elimination as the main implementation levers, while Gunasekaran et al. (1993) described JIT as a manufacturing-system approach that reaches across business functions. Shah and Ward (2003) showed that lean practices deliver results as bundles and not as isolated tools. In a meta-analysis, MacKelprang and Nair (2010) found a positive overall association between JIT practices and performance, although the strength varied from practice to practice. Yavuz and Akçali (2007) reviewed the production-smoothing models that support JIT, and Mahendran et al. (2018) documented a lean implementation in an automobile plant.

    2. Integrated Planning and Battery Supply Chains

      Xu and Song (2022) modelled raw-material ordering, production planning, and capacity together under uncertain lead times and quantities. For batteries specifically, Zou et al. (2025) optimised production and inventory capacities under demand uncertainty, Dhairiyasamy et al. (2024) identified import dependence, fragmented supply chains, production cost, quality control, and rapid technological change as the main barriers in Indias EV-battery supply chain, and Yang et al. (2026) examined battery production decisions and supplier cooperation. Standard operations texts (Chopra, 2019; Christopher, 2016; Heizer et al., 2017; Slack et al., 2022) supply the BOM, capacity, and inventory-control foundations, and Silver et al. (2017) provide the inventory-control models on which reorder logic rests.

    3. Perishability, Variability, and Resilience

      Classical perishable-inventory work shows that ageing stock changes the optimal ordering policy (Ghare & Schrader, 1963; Nahmias, 1982). Research on the bullwhip effect shows that order variability is amplified as it moves upstream (Lee et al., 1997), and resilience studies caution that lean supply chains concentrate disruption risk (Christopher & Peck, 2004). Together these streams suggest that plan volatility, shelf life, and lead time should be analysed jointly, which the JIT literature seldom does.

    4. Research Gap

      The literature is strong on JIT, lean systems, and integrated planning, but it says little about shelf-life- sensitive adhesives, FEFO, design-change obsolescence, and customer-rejected packs in

      battery-pack assembly, and it rarely links daily PPC achievement to expiry exposure. Methodologically, few industrial case studies reconcile small, availability-limited samples with inferential modelling or report data-quality diagnostics. This study addresses both gaps.

  4. RESEARCH OBJECTIVES, CONCEPTUAL FRAMEWORK, AND HYPOTHESES

    1. Research Objectives

      The study pursues four objectives. O1: to compute BOM-level material requirements for 10,000 battery packs per month from rated capacity and explicit planning assumptions. O2: to design supplier ordering and JIT delivery that reflect a 2% rejection allowance, 1% minimum stock, a 10% supplier- order increase, and supplier rejection. O3: to assess transport cost, floor utilisation, floor-space cost, and an illustrative reduction in inventory value. O4: to bring PPC performance, imported-material lead time, shelf-life exposure, and obsolete and customer-rejected inventory into one control framework.

    2. Conceptual Framework and Conceptual Novelty

      Figure 1 shows the framework. Demand, capacity, and the BOM feed a planning layer of buffers, supplier orders, and controlled JIT. Execution is governed by PPC plan-versus-actual feedback, FEFO issue, and import lead time. A special- inventory loop and a set of new decision constructs sit beneath this flow and discipline it. The conceptual novelty lies in these constructs (Table 1), which convert qualitative good practice into numbers that a planner can compute, monitor, and act upon.

      Figure 1

      Integrated Framework for Controlled JIT and Inventory Control

      Note. Constructs in the lower row are original to this study. Arrows show the direction of influence.

      ID

      Obj.

      Hypothesis

      Test

      H2

      O2

      The probability of a material shortfall rises with supplier rejection and demand volatility and falls with the supplier-order uplift.

      M2 (logit)

      H3

      O3

      Net monthly JIT benefit rises with the carrying-cost rate and inventory reduction and falls with delivery frequency and cost per trip.

      M3 (OLS)

      H4

      O4

      The probability of expiry risk falls as PPC achievement rises and rises with days of cover; a lot is safe only when A

      A*.

      M4 (logit)

      H5

      O4

      Imported-material lead time exceeds coverage (LCG > 0).

      Descriptive

      H6

      O4

      An engineering- change gate and an RCA loop raise SICR and lower repeat rejection.

      Needs post- implementation data

      Table 1

      Novel Decision Constructs Introduced in This Study

      27.2%

      (base case)

      Construct

      Definition and formula

      Value in this study

      Expiry-safe achievement threshold (A*)

      Minimum average PPC achievement at which a lot is consumed before expiry: S ÷ (D × u)

      3.8%

      35.3%

      (Table 12)

      Buffer- stacking inflation (BSI)

      Compounded uplift from sequential buffers: (1 + r)(1 + m)(1 + s) 1

      13.3%

      JIT break- even carrying rate (c*)

      Annual carrying-cost rate at which JIT breaks even: 12 × (T F) ÷ I

      Lead-time coverage gap (LCG)

      Total lead time minus days of coverage

      10 days

      Special- inventory closure rate (SICR)

      Units dispositioned ÷ (obsolete + customer- rejected units)

      650 units to close

      Controlled JIT

      eligibility matrix

      Classification of materials by source (local or import) and shelf-life sensitivity

      Table 20

      Note. S = stock; D = days to expiry; u = planned daily use; r, m, and s = rejection, minimum-stock, and supplier-order uplifts; T, F, and I = monthly transport increase, monthly floor saving, and inventory-value reduction.

    3. Hypotheses

      Each objective is paired with at least one testable hypothesis (Table 2). Hypotheses H1H4 are tested through the regression models in Section 7, H5 is tested descriptively, and H6 requires post- implementation data that do not yet exist.

      Table 2

      Research Hypotheses

      ID

      Obj.

      Hypothesis

      Test

      H1

      O1

      Under plan-only ordering, overstock falls as PPC achievement rises, and achievement

      has a larger standardised effect

      than any buffer.

      M1 (OLS)

  5. METHODOLOGY

    1. Research Design and Data Collection

      An applied, descriptive and analytical single-case study was carried out during on-the-job training from 18 May to 31 August 2026 in Stores and PPC. The unit of analysis is one battery pack and its BOM components over a 26-working-day horizon. Five inputs were used: (a) the planning baseline of 10,000 packs per month, 26 working days, two shifts, 350 packs per shift, and 18,000 packs of rated monthly capacity; (b) PPC daily records for 2430 July and 410 August 2026, twelve dates in all, with 31 July

      3 August unavailable and deliberately not reconstructed; (c) a stores snapshot of shelf-life- sensitive materials dated 25 August 2026; (d) plant- supplied counts of 400 obsolete and 250 customer- rejected packs; and (e) hands-on exposure to ERP/SAP stock checks, goods receipt, material issue, and supplier follow-up (Endurance Technologies Ltd., New Energy Plant, 2026).

    2. Sampling Methodology

      The population is every material-flow event of the battery-pack line during the training period, and the sampling frame is the set of PPC daily reports, stores records, and ERP transactions to which the author had approved access. Sampling was non-probability and purposive. A single plant was chosen as a critical case, because case logic aims at analytic rather than statistical generalisation (Flyvbjerg, 2006; Yin, 2018). Within the case, PPC dates were taken from two consecutive calendar weeks, each containing six working days, because these were the records available and verifiable. The shelf-life- sensitive materials and the special inventory were examined as complete censuses (Table 3).

      Table 3

      Element

      Sampling design

      Size

      Plant

      Purposive, single critical case

      1 plant

      PPC

      records

      Availability- constrained, two consecutive-week blocks

      12 working

      dates; 31 Jul

      3 Aug

      missing, not imputed

      Shelf-life materials

      Census of the 25 August 2026

      snapshot

      6 material lines

      Special inventory

      Census

      650 packs

      (400 + 250)

      Regression inputs

      Monte Carlo draws calibrated to observed parameters

      5,000 monthly

      cycles; 4,998 lot draws

      Sampling Design by Data Element

      gap awareness treats missing dates as potentially informative (Rubin, 1976): they are logged, never imputed, and bounded through best-case and worst- case sensitivity checks. Third, dual-ledger matching links each production date to the stores snapshot and the expiry ledger, which is what makes A* computable. Fourth, event anchoring over- samples disruption days (power failure, PDI errors, clearance delays) and expiry-proximate lots, because these carry most of the managerial information. In this paper the first three features are applied in a limited form; the fourth is specified for the extension.

        1. Data Bias and Genuineness

          Because the paper rests on internal records gathered by a participant observer, bias and authenticity are treated explicitly. Table 4 lists the main biases, their likely effects, and the mitigations applied. In addition, the arithmetic of every plant table (Tables 614) was independently recomputed during this revision and reconciled with the stated inputs.

          Table 4

          Note. Simulation draws are synthetic and are not field observations.

          A sample of twelve dates cannot support multivariate regression. Greens (1991) rule of N

          50 + 8k implies about 90 observations for five predictors. For this reason the regression models in Section 7 are calibrated by simulation, and the study specifies a field-collection protocol for later re- estimation: at least 90 consecutive production days and at least 30 shelf-life lots, each with a matched PPC date and stores snapshot.

    3. Sampling Novelty

      The study proposes event-anchored, gap-aware, dual-ledger (EGD) sampling, designed for small industrial cases in which records are patchy and disruptions drive outcomes. It has four features. First, block sampling keeps consecutive days together so that autocorrelation and disruption cascades are preserved, and uncertainty is expressed with a moving-block bootstrap (Künsch, 1989) rather than an independence assumption. Second,

      Bias Register

      Bias

      Source in this study

      Mitigation

      Selection

      Twelve dates in two blocks; June has no verified record; four dates missing

      Report as

      sample, not month totals; no imputation; sensitivity bounds

      Disruption- period

      Sample window contains many power and machine disruptions

      Event- anchored sampling; extend to 90+ days

      Participant- observer

      Intern embedded in Stores and PPC

      Mentor review; triangulation with ERP/SAP

      records

      Self-reported reasons

      Gap reasons entered by shift staff; MES data sometimes not cleared

      Cross-check with dispatch and MES; standard reason codes

      Recording

      Two expiry dates unverified,

      including one Henkel lot

      Flag; verify against physical lot labels

      Datas et

      Pro v.

      Ver if.

      Com pl.

      Consi st.

      Time ly

      Tot al

      Import lead- time map

      1

      1

      2

      2

      1

      7

      Cost, transp ort, and floor scenari os

      0

      0

      1

      2

      1

      4

      Bias

      Source in this study

      Mitigation

      Anonymisation

      Sensitive values anonymised, illustrative, or forecast

      Label as directional; replace with actual

      Planning-rate mismatch

      Daily PPC plans of 150300

      versus the 400- pack target

      Recompute A* at actual plan; report both

      Synthetic data

      Regression inputs are simulated

      Label as model-based; use wide

      ranges; re-

      estimate on field data

      Confirmation

      Prior preference for controlled JIT

      Report null results (r and m); apply the c* break-even test

      Single site

      One plant

      Analytic generalisation only (Yin, 2018)

      Genuineness is assessed with a five-criterion Data Genuineness Index (DGI): provenance, verifiability, completeness, consistency, and timeliness, each scored 0 (poor), 1 (partial), or 2 (strong), for a maximum of 10. Table 5 gives a provisional rating derived solely from the limitations stated in this paper. Datasets scoring 7 or more are treated as decision-grade for the descriptive tables, and the cost scenario, at 4, is treated as directional only. The author should confirm these ratings against the source documents.

      Table 5

      Datas et

      Pro v.

      Ver if.

      Com pl.

      Consi st.

      Time ly

      Tot al

      PPC

      daily record s

      2

      1

      1

      1

      2

      7

      Stores snapsh ot (25 Aug 2026)

      2

      1

      1

      2

      2

      8

      Specia l- invent ory

      counts

      2

      1

      1

      2

      2

      8

      Data Genuineness Index (Provisional)

      Note. Prov. = provenance; Verif. = verifiability; Compl. = completeness; Consist. = consistency; Timely = timeliness. Scores are 02 per criterion.

        1. Analytical Framework and Limitations

      Material requirements follow a fixed sequence. The basic requirement is monthly volume × BOM quantity per pack; this is then multiplied by 1.02 for rejection, 1.01 for minimum stock, and 1.10 for the supplier-order increase, and a separate supplier- rejection adjustment follows. For shelf-life items, daily consumption = daily plan × quantity per pack; coverage days = stock ÷ daily consumption; days to expiry = expiry date planning date; and expiry-risk quantity = max(0, stock days to expiry × daily consumption). Production achievement = actual ÷ plan × 100. The expiry-safe threshold follows directly: a lot is consumed in time when achievement is at least A* = S ÷ (D × u). Regression strategy is described in Section 7.

      Several limitations apply. The cost, transport, and floor-space figures are illustrative. Actual supplier- rejection rates, MOQ, and adhesive lead times were not supplied. The PPC sample covers only twelve dates, and its daily plans (150300 packs) differ from the 400-pack operational target. Two expiry dates are unverified (one Henkel glue lot and one 40 kg sealant lot). No post-implementation trial was measured, so results are model-based. Finally, the regression coefficients reflect the assumed simulation ranges and must not be read as field estimates.

  6. DATA ANALYSIS AND DISCUSSION

    1. Converting Demand into Capacity and Material Requirements

      Table 6

      Planning Baseline and Capacity

      Parameter

      Basis

      Value

      Monthly demand

      Given

      10,000 packs

      Average daily demand

      10,000 ÷ 26

      384.6 packs

      Parameter

      Basis

      Value

      Operational daily target

      Rounded

      400 packs

      Plan per shift (2 shifts)

      400 ÷ 2

      200 packs

      Theoretical capacity per day

      350 × 2

      700 packs

      Rated monthly capacity

      Given

      18,000 packs

      Monthly demand utilisation

      10,000 ÷

      18,000

      55.56%

      Target utilisation

      400 ÷ 700

      57.14%

      Rated monthly headroom

      18,000

      10,000

      8,000 packs

      coverage across two eight-hour shifts would strictly imply four deliveries per day; with two, each delivery covers a full shift of about 16,000 basic units. Four daily deliveries would double the trip count to 104, or 5.2 lakh per month at the assumed

      5,000 per trip. Delivery frequency is therefore a logistics assumption to be tested against MOQ and vehicle capacity.

      Table 8

      Measure

      Before

      After

      Change

      Trips per month

      4

      52

      +48

      Transport cost per month

      20,000

      2,60,000

      +2.40

      lakh

      Inventory equivalent

      5,000

      packs

      1,500

      packs

      70%

      Floor area (0.5 sq ft per pack)

      2,500 sq

      ft

      750 sq ft

      1,750

      sq ft

      Floor cost (40 per sq ft)

      1,00,000

      30,000

      70,000

      Inventory value

      4.00 Cr

      3.25 Cr

      18.75%

      Illustrative Transport, Floor-Space, and Inventory Scenarios

      Table 6 keeps the 10,000-pack monthly baseline separate from the 400-pack shop-floor target, so that a demand assumption is not confused with a rounded operating figure. The plant uses just over half of its rated monthly capacity, so capacity is not the binding constraint; material availability and plan adherence are (Proposition 3).

      Table 7

      Illustrative BOM Calculation (80 Units per Pack)

      Step

      Calculation

      Quantity

      Basic requirement

      10,000 × 80

      800,000

      After 2%

      rejection

      800,000 ×

      1.02

      816,000

      After 1%

      minimum stock

      816,000 ×

      1.01

      824,160

      Supplier order (+10%)

      824,160 ×

      1.10

      906,576

      Per delivery (52 per month)

      906,576 ÷ 52

      17,434

      The three buffers lift the supplier order 13.3% above basic need (BSI = 13.3%). The 10% supplier-order increase is applied after the rejection and minimum- stock steps, and supplier rejection is a separate quality adjustment whose actual percentage must come from approved supplier data. A uniform 13.3% uplift sits uneasily beside an inventory- reduction goal, so buffers should be justified by material class, lead time, and supplier performance and not applied to every part.

    2. Controlled JIT and Total Cost

      At 200 packs per eight-hour shift the line consumes 25 packs per hour, so four-hour line-side coverage equals 100 packs, or 8,000 units for the 80-unit example BOM. The delivery model assumes two deliveries per day, or 52 per month. Four-hour

      Note. All figures are illustrative scenarios supplied fo planning.

      Table 8 shows the central trade-off. JIT releases 1,750 sq ft and saves 70,000 per month in floor cost (8.40 lakh annually) and 0.75 crore of inventory value, but adds 2.40 lakh per month in transport. The scenarios do not complete the total-cost comparison, so the break-even rate c* is derived here (Proposition 1). After netting the floor saving, the carrying saving on 0.75 crore must exceed

      1.70 lakh per month, equivalent to c* = 27.2% per year. At ordinary carrying-cost rates the scenario is cost-negative, which supports controlled JIT for suitable local materials with consolidated deliveries where MOQ and vehicle capacity allow.

      Delivery frequency changes the answer sharply. Holding all other inputs constant, c* is 6.4% per year at 26 trips per month (one delivery a day), 27.2% at 52 trips (two a day), and 68.8% at 104 trips (four a day). A four-delivery-a-day regime would therefore be uneconomic at any realistic carrying-cost rate, whereas a daily-delivery regime pays for itself at a modest rate (Figure 2). The 70% unit reduction and the 18.75% value reduction are separate illustrations and should be reconciled with plant data.

      Figure 2

      Net Monthly JIT Benefit by Carrying-Cost Rate and Delivery Frequency

      Note. Drawn from the inputs in Table 8 at 5,000 per trip. Lines cross zero at c* = 6.4%, 27.2%, and 68.8%.

    3. Imported-Material Lead-Time Exposure

      Table 9

      Import Lead-Time Map

      Element

      Days

      Element

      Days

      Shipping China to India

      21

      Plant movement

      5

      China customs

      7

      Quality clearance

      5

      India customs

      7

      Total lead time

      55

      Truck movement

      10

      Project coverage / gap

      45 /

      10

      Table 9 shows a 55-day end-to-end lead time against a 45-day coverage assumption, a lead-time coverage gap of 10 days (H5 supported). Imported and long- lead materials therefore fall outside the four-hour JIT window and need lead-time-based protection: a shorter lead time, supplier-held stock, or a revised safety stock. The exposure is consistent with the import dependence reported for Indias EV-battery supply chain (Dhairiyasamy et al., 2024).

    4. Plan-versus-Actual Performance through PPC

      complete-month result. Daily achievement ranged from 0% on 25 July to 169.8% on 27 July, and fell

      to 13.6% and 11.1% on 29 and 30 July. Dispatch was zero on 24, 29, and 30 July against a daily plan of 162, because of pending customer quality clearance and data not cleared in the manufacturing execution system (MES).

      The recorded reasons for gaps are varied: casting and rework availability, an ALT machine or PLC breakdown, high delta-voltage failures at the CDC stage, a power failure and a transformer fault at the CDC area, repeated power cuts, bootloader errors at pre-delivery inspection (PDI), and clearance delays. The PPC report is therefore a cross-functional control document linking Production, Quality, PDI, Dispatch, Engineering, and the customer. It also feeds inventory: when output falls below plan, consumption falls and shelf-life exposure rises, and when output recovers, short-coverage materials need faster replenishment. A daily PPCStores review of actual output, coverage days, lead time, and FEFO status is the recommended control.

    5. Shelf-Life-Based Control of Cell Glue and Sealants

      The earlier approach ordered adhesives by quantity, without linking orders to the latest PPC plan or to remaining shelf life. The redesigned method converts stock into pack coverage and coverage days, checks days to expiry before any purchase order, issues the earliest-expiry batch first, plans Cell Glue Part A and Part B together because both are needed, and sets the next order from lead-time demand plus minimum stock less usable stock.

      Table 11

      Shelf-Life Coverage at 400 Packs per Day (25 August 2026)

      Material

      Stock

      Daily use

      Cover (days)

      Earliest expiry

      Cell glue, Henkel

      72 L

      4 L

      18.00

      15-Oct-

      26*

      Cell glue, Part A

      45 L

      4 L

      11.25

      15-Oct-

      26

      Cell glue, Part B

      45 L

      4 L

      11.25

      15-Oct-

      26

      Loctite (screws)

      77,750

      mL

      4,000

      mL

      19.44

      24-Oct-

      26

      Sealing sealant

      820 kg

      18.4

      kg

      44.57

      10-Feb-

      27

      Connector silicon sealant

      64,800

      mL

      6,800

      mL

      9.53

      30-Apr-

      27

      Table 10

      Observed PPC Achievement on Recorded Dates

      Month

      Dates

      Plan

      Actual

      Achievement

      July 2026

      6

      972

      494

      50.82%

      August 2026

      6

      1,650

      1,149

      69.64%

      Note. Dates: 2430 July and 410 August 2026. These are sample totals, not complete-month totals; June has no verified record.

      Achievement rose by 18.82 percentage points between the two samples. The improvement is real for the recorded dates but cannot be read as a

      Note. *Henkel expiry is not recorded in the source sheet and must be verified. One 40 kg lot has no expiry recorded. Expiry-risk quantity is 0 for all dated lots at this rate.

      At 400 packs per day no dated lot carries a positive expiry-risk quantity, because each lot is consumed well before expiry. The result is rate-dependent, however: PPC achievement was at or below 92.3% of plan on eleven of the twelve recorded dates, and the same stock would then take longer to consume. The redesigned system therefore recalculates risk from actual output each cycle. The comparison with the earlier scrap-quantity field is a model-based comparison at the stated planning rate, not a measured post-implementation wastage trial.

      The expiry-safe threshold A* makes this rate- dependence explicit (Proposition 2). Table 12 shows that A* ranges from 3.8% for connector silicon sealant to 35.3% for Henkel glue. Every threshold lies far below the observed 50.82% and 69.64% (Figure 3), which explains why no expiry-risk quantity appears. The binding materials are Henkel glue and Loctite: if average achievement over the remaining shelf life stayed below about one-third of plan, stock would be stranded. Single-day dips to 0%, 13.6%, and 11.1% matter far less for expiry than for line stoppage, because lots are consumed cumulatively and only the horizon average must clear A*.

      Table 12

      Expiry-Safe Achievement Thresholds

      Material

      Cover (days)

      Days to expiry

      A* (%)

      Model P(A < A*) (%)

      Cell glue, Henkel

      18.00

      51

      35.3

      4.7

      Cell glue, Part A

      11.25

      51

      22.1

      0.4

      Cell glue, Part B

      11.25

      51

      22.1

      0.4

      Loctite (screws)

      19.44

      <>60

      32.4

      3.1

      Sealing sealant

      44.57

      169

      26.4

      1.1

      Connector silicon sealant

      9.53

      248

      3.8

      < 0.1

      Note. Days to expiry are counted from 25 August 2026. A* = S ÷ (D × u) at a 400-pack daily plan. Model probabilities assume achievement follows a Beta distribution with mean 0.62 and SD 0.15 (Table 16). The Henkel expiry date is unverified.

      Figure 3

      Expiry-Safe Achievement Thresholds Compared With Observed PPC Achievement

      Note. Bars show A*. Vertical lines show observed achievement on the recorded July and August dates.

      A* guards only the lower side of a shelf-life corridor. Connector silicon sealant has the lowest A* but also the shortest cover, 9.53 days, so its exposure is to shortage and not to expiry. The upper side of the corridor is the achievement level at which stock runs out before the next delivery arrives, S ÷ (u × L) for supplier lead time L, which cannot be computed here because adhesive lead times were not supplied. Collecting those lead times would complete the corridor for every material.

      Table 13

      No-Replenishment Shortage Risk, September November 2026

      Material

      Monthly demand

      Sep deficit

      3-month deficit

      Cell glue, Henkel

      100 L

      28 L

      228 L

      Cell glue, Part A / B

      100 L

      each

      55 L

      255 L

      each

      Loctite

      100,000

      mL

      22,250

      mL

      222,250

      mL

      Sealing sealant

      460 kg

      0 kg

      560 kg

      Connector silicon sealant

      170,000

      mL

      105,200

      mL

      445,200

      mL

      Note. Theoretical deficits if no inward supply is received and 10,000 packs per month is maintained. These are risk indicators, not forecasts of actual shortage.

      Table 13 sizes the replenishment task. Opening stock covers September demand for sealing sealant but falls short for the other families from the first month, so ordering must continue under the lead- time-demand rule, with FEFO and shelf-life checks, rather than move toward zero-inventory purchasing.

      Recommendation

      Expected outcome

      KPI

      Supplier monitoring

      Lower shortage risk

      Supplier rejection %; on-time delivery

      Inventory and floor dashboard

      Lower capital blockage; space release

      Inventory value; floor utilisation

      Import lead-time review

      Protection against supply interruption

      LCG (days)

      Daily PPCStores feedback with A* alert

      Faster corrective action

      Coverage days; A*; expiry-risk quantity

      Engineering- change gate

      Lower obsolete inventory

      Obsolete quantity; SICR

      Rejection RCA dashboard

      Lower repeat rejection

      Repeat- rejection rate

      5S and FEFO discipline

      Lower expiry and identification risk

      Expiry scrap quantity; audit score

    6. Obsolete and Customer-Rejected Inventory

      Table 14

      Special Inventory Categories and Controls

      Category

      Qty

      Driver

      Control sequence

      Obsolete

      400

      Customer design change at SOP

      Segregate; verify approved design; controlled dismantling; recover approved cells and

      components after inspection; powder- coat approved covers; scrap the rest under authority; retest; close in ERP

      Customer rejected

      250

      Customer quality rejection

      Segregate; defect analysis; RCA; rework feasibility; rework and retest; customer approval and re-dispatch, or recovery or scrap

      Total

      650

      Controlled closure

      The two categories need different controls because their causes differ. Obsolescence is an engineering- change failure, so the preventive control is an inventory gate before SOP: when a customer design change is approved, PPC, Engineering, Stores, Production, and Quality jointly identify and freeze pre-change finished goods, work in process, and critical components. Customer rejection is a quality failure, so the control is a closed loop that prevents repeat defects. Progress can be tracked through obsolete quantity, recovery rate, and SICR for the first category, and rejection rate, repeat-rejection rate, rework recovery rate, and SICR for the second. Defect-wise data for the 250 rejected packs were not supplied, so root-cause categories could not be analysed.

    7. Recommended Improvements

      Table 15

      Recommendation

      Expected outcome

      KPI

      Controlled JIT for suitable local parts, with c* test

      Lower line- side excess stock

      Delivery- frequency adherence; c* versus actual

      carrying rate

      Recommendations, Expected Outcomes, and KPIs

  7. REGRESSION ANALYSIS

    1. Rationale and Simulation Design

      Twelve PPC dates and six material lines cannot support inferential regression, and several drivers (supplier rejection, carrying cost, adhesive lead time) were not supplied. Each objective is therefore expressed as a regression on a calibrated Monte Carlo process (Table 16). Inputs are anchored to observed plant values where these exist and otherwise take wide, plausible ranges, in the spirit of global sensitivity analysis (Saltelli et al., 2008). The resulting data are synthetic. The regressions recover the relationships implied by the planning logic and show which inputs dominate; they are not field evidence, and their coefficients depend on the assumed ranges. Their value is to prioritise what to measure and to fix an analysis plan before real data arrive.

      Table 16

      Simulation Parameters

      Variable

      Distribution

      Anchor

      Monthly demand, V

      Normal (10,000;

      1,000), truncated

      to 6,00014,000

      Planning baseline

      Variable

      Distribution

      Anchor

      PPC

      achievement, A

      Beta, mean 0.62,

      SD 0.15

      Observed

      0.508 (July)

      and 0.696 (August)

      Rejection allowance, r

      Uniform 1%3%

      2% in Table

      7

      Minimum stock, m

      Uniform 0.5%

      1.5%

      1% in Table

      7

      Supplier- order uplift, s

      Uniform 5%15%

      10% in Table

      7

      Supplier rejection,

      Uniform 0%12%

      Assumed (not supplied)

      Demand volatility, CV

      Uniform 5%25%

      Assumed

      Carrying- cost rate, c

      Uniform 12%

      30% per year

      Assumed

      Cost per trip, t

      Uniform 3500

      6,500

      5,000 in Table 8

      Trips per month, n

      26, 52, or 104 with equal probability

      Daily, twice daily, four times daily

      Floor rate, f

      Uniform 3050 per sq ft

      40 in Table 8

      Inventory reduction, I

      Uniform 50%

      100% of 0.75

      crore

      Table 8

      Lot stock (M4)

      Observed stock × Uniform (0.6, 1.6)

      Table 11

      and logit models by maximum likelihood. A bootstrap interval (Efron, 1979; 300 resamples) was computed for the carrying-rate coefficient in M3. Inputs were drawn independently, so multicollinearity is absent by construction.

      7.3 Results

      Table 17

      Predictor

      b

      SE

      p

      Panel A. M1: ln(1 + overstock ratio); N = 5,000; R² = .966

      Intercept

      1.651

      0.063

      <

      .001

      PPC

      achievement, A (01)

      1.776

      0.010

      .978

      <

      .001

      Rejection allowance, r

      0.865

      0.115

      .019

      <

      .001

      Minimum stock, m

      1.187

      0.242

      .013

      <

      .001

      Supplier-order uplift, s

      0.882

      0.025

      .094

      <

      .001

      ln(Monthly demand)

      0.003

      0.007

      .001

      .622

      Panel B. M3: net monthly JIT benefit ( lakh); N

      = 5,000; R² = .974

      Intercept

      2.0604

      0.0518

      <

      .001

      Carrying-cost rate (percentage points)

      0.0465

      0.0008

      .137

      <

      .001

      Cost per trip ( thousand)

      0.5722

      0.0062

      .284

      <

      .001

      Trips per month

      0.0500

      0.0002

      .921

      <

      .001

      Floor rate ( per sq ft)

      0.0175

      0.0007

      .058

      <

      .001

      Inventory reduction ( crore)

      1.7898

      0.0379

      .110

      <

      .001

      OLS Regression Results (HC3 Robust Standard Errors)

      Note. Seed = 2026; 5,000 monthly cycles for M1 M3 and 4,998 lot draws for M4.

    2. Model Specification

      M1 (O1, requirement planning). ln(1 + OS) = 0

      + 1A + 2r + 3m + 4s + 5 ln V + , where overstock OS = Q ÷ (80 × V × A) 1 and Q = 80 × V × (1 + r)(1 + m)(1 + s) is the plan-based order.

      M2 (O2, supply continuity). logit P(Shortfall = 1)

      = 0 + 1 + 2CV + 3r + 4m + 5s, where shortfall occurs when Q × (1 ) is less than the actual need 80 × V* × (1 + r*), with V* drawn from Normal(V, CV × V) and r* from Uniform(0.5r, 1.5r).

      M3 (O3, total cost). Net = 0 + 1c + 2t + 3n + 4f + 5I + , where Net is the floor saving plus the carrying-cost saving less the added transport cost ( lakh per month).

      M4 (O4, shelf-life control). logit P(Expiry risk = 1)

      = 0 + 1A + 2Cover + 3D, where expiry risk occurs when S D × u × A > 0.

      Linear models were estimated by ordinary least squares with HC3 heteroskedasticity-robust standard errors (White, 1980; Wooldridge, 2020),

      Note. b = unstandardised coefficient; = standardised coefficient. Results are simulation- based.

      Model M1. Plan adherence dominates. A ten-point rise in PPC achievement lowers the log overstock ratio by 0.178, roughly a 16% lower ratio, and its standardised effect ( = .978) is about ten times that of the largest buffer (supplier-order uplift, =

      .094). Under plan-only ordering at the assumed achievement distribution, mean overstock is 96.6% and median overstock is 81.1%, consistent with

      Proposition 4. Monthly demand has no independent effect (p = .622). H1 is supported within the model. Model M3. Mean net JIT benefit across scenarios is

      1.15 lakh per month, and only 35.1% of scenarios are positive. Each additional percentage point of carrying cost adds 0.047 lakh per month (95% bootstrap interval 0.045 to 0.048), but delivery frequency dominates ( = .921): every extra monthly trip costs 0.05 lakh. H3 is supported within the model, and the finding matches the analytical break-even in Section 6.2.

      Table 18

      Logistic Regression Results

      Predictor

      b

      SE

      Odds ratio

      p

      Panel C. M2: material shortfall (1 = yes); N = 5,000; prevalence = 36.7%; McFadden R² = .064

      Intercept

      1.066

      0.209

      0.345

      <

      .001

      Supplier rejection, (pp)

      0.139

      0.009

      1.149

      <

      .001

      Demand volatility, CV (pp)

      0.045

      0.005

      1.046

      <

      .001

      Rejection allowance, r (pp)

      0.056

      0.053

      1.057

      .293

      Minimum stock, m (pp)

      0.046

      0.106

      0.955

      .663

      Supplier-order uplift, s (pp)

      0.110

      0.011

      0.896

      <

      .001

      Panel D. M4: expiry risk (1 = yes); N = 4,998; prevalence = 3.8%; McFadden R² = .895

      Intercept

      28.799

      3.272

      <

      .001

      PPC achievement, A (pp)

      0.748

      0.083

      0.473

      <

      .001

      Days of cover

      1.247

      0.146

      3.480

      <

      .001

      Days to expiry

      0.483

      0.057

      0.617

      <

      .001

      Note. pp = percentage points. Odds ratios in Panel D are inflated by near-perfect separation and should be read for direction and threshold behaviour only.

      Model M2. Shortfall occurs in 36.7% of cycles when orders are sized to plan. Each percentage point of supplier rejection raises the odds of a shortfall by 14.9%, each point of demand volatility by 4.6%, and each point of supplier-order uplift lowers them by 10.4%. The 2% rejection allowance and 1% minimum stock are not significant (p = .293 and

      .663), which suggests that, within the sampled ranges, they add little protection on their own; the 10% uplift is the effective shock absorber. H2 is therefore partly supported. The low pseudo-R²

      (.064) reflects the large random component in demand and rejection.

      Model M4. Expiry risk is a threshold phenomenon. Higher achievement lowers its odds (odds ratio 0.473 per point), whereas more days of cover raise them (3.48 per day) and longer remaining shelf life lowers them (0.617 per day), as Proposition 2 implies. The very high pseudo-R² (.895) shows the near-deterministic nature of the threshold and not a rich behavioural fit. Using Table 12, the model probability that average achievement falls below A* is .7% for Henkel glue, 3.1% for Loctite, 1.1% for sealing sealant, 0.4% for Part A and Part B, and below 0.1% for connector silicon sealant. H4 is supported within the model.

        1. Hypothesis Verdicts

          Table 19

          Summary of Hypothesis Tests

          ID

          Verdict

          Key evidence

          H1

          Supported (model- based)

          for A = .978 versus for s

          = .094 (Table 17)

          H2

          Partly supported (model- based)

          , CV, and s significant; r (p =

          .293) and m (p = .663) not

          H3

          Supported (model- based)

          Mean net benefit 1.15 lakh; 35.1% of scenarios positive; c* = 27.2% in the base case

          H4

          Supported (model- based)

          A and D lower the odds of expiry risk; cover raises them; P(A < A*) = 0.0%4.7%

          H5

          Supported (descriptive)

          LCG = 10 days (Table 9)

          H6

          Not yet testable

          Needs post-implementation SICR and rejection data

        2. Robustness and Caveats

      Signs and relative magnitudes follow from the planning logic and are therefore stable, but the coefficient sizes depend on the assumed ranges, and no further robustness variants were run. The simulation should be re-estimated on field data collected under the protocol in Section 5.2, and the departure of the field estimates from these model- based values would itself be informative about hidden frictions in the plant.

  8. CONTRIBUTIONS OF THE STUDY

    1. Theoretical Contributions

      The paper moves beyond the claim that JIT is beneficial by stating the conditions under which it is not. Proposition 1 gives an explicit economic threshold for JIT (c*), Proposition 2 turns expiry

      from an inventory-age problem into a plan- adherence problem, Proposition 3 applies the theory of constraints to show that, at moderate utilisation, the constraint moves from capacity to material and plan reliability, and Proposition 4 quantifies how plan-blind buffers become overstock. Together they link pull logic, perishable-inventory theory, and constraint thinking in one analytical frame, which the existing literature treats separately.

    2. Conceptual Contributions

      Five measurable constructs (A*, BSI, c*, LCG, and SICR) and a classification device, the controlled JIT eligibility matrix (Table 20), convert qualitative good practice into numbers a planner can compute from existing Stores and PPC data. The matrix replaces the all-or-nothing question of whether to adopt JIT with a material-by-material decision.

      Table 20

      Material class

      Local (short lead time)

      Imported (long lead time)

      Shelf-life- insensitive

      Pull JIT. Frequent small deliveries, with the delivery frequency chosen so that c* stays below the actual carrying rate.

      Lead-time protection. Coverage at least equal to lead time; supplier- held stock; work to close LCG.

      Shelf-life- sensitive

      FEFO-pulled JIT. Order

      against actual output; monitor A*; plan paired lots (Part A and Part B) together.

      Dual protection. Smaller, staggered shipments; FEFO; monitor both A* and LCG.

      Controlled JIT Eligibility Matrix

      A phased roll-out is feasible. In the first 30 days, compute A*, LCG, and coverage for all shelf-life and imported materials and fix the data gaps in Section 5.4. In days 31 to 60, launch the daily PPC Stores loop and the rejection RCA dashboard. In days 61 to 90, apply the eligibility matrix, pilot controlled JIT on two or three local parts, and begin collecting the field data needed to re-estimate the regressions. Taken together these elements form a plant control tower in which four dashboards (coverage and expiry, lead time, cost, and special- inventory closure) share one daily data feed.

      8.5 Industry, Policy, and Sustainability Contributions

      The study supports the localisation goals behind PM E-DRIVE (MHI, n.d.) by showing how a domestic assembler can protect continuity without bloating inventory. Its recovery-focused handling of 650 obsolete and rejected packs, which recovers approved cells and components instead of defaulting to scrap, aligns with the extended-producer- responsibility logic of the Battery Waste Management Rules, 2022 (Ministry of Environment, Forest and Climate Change, 2022). Lower expiry scrap and lower blocked floor space also reduce material waste and the physical footprint of the plant.

  9. FINDINGS, CONCLUSIONS, AND SUGGESTIONS

    8.3 Methodological Contributions

    Three methodological contributions are offered. EGD sampling gives small, patchy industrial case studies a principled way to handle blocks, gaps, and disruption days. Simulation-calibrated regression shows how a case with too few observations can still state pre-specified, falsifiable models and identify the variables worth measuring. The Data Genuineness Index gives practitioners a short, auditable scorecard that makes data quality visible alongside results.

    8.4 Managerial and Practical Contributions

    For the plant, the paper offers a decision toolkit that can be run from existing records: an A* alert in the daily PPCStores review, a c* test before any increase in delivery frequency, a buffer budget by material class in place of a uniform 13.3% uplift, and an engineering-change inventory gate before SOP.

    1. Findings

      The baseline uses 55.56% of the 18,000-pack rated monthly capacity, and the 400-pack operational target uses 57.14% of theoretical daily capacity. A four-hour JIT window equals 100 packs. Stacked buffers raise the illustrative supplier order from 800,000 to 906,576 units. In the illustrative scenario JIT releases 1,750 sq ft and 70,000 per month of floor cost and lowers inventory value by 18.75%, but adds 2.40 lakh per month in transport; the break-even carrying rate is 27.2% at two deliveries a day and ranges from 6.4% to 68.8% across delivery frequencies. Imported material shows a 10- day exposure. Observed PPC achievement was 50.82% in July and 69.64% in August across the recorded dates, held back by power, machine, PDI, and clearance disruptions. At 400 packs per day no dated adhesive lot carries expiry risk, coverage ranges from 9.5 to 44.6 days, and expiry-safe thresholds range from 3.8% to 35.3% of plan. In the simulation-calibrated models, plan adherence outweighs buffer size as a driver of overstock, delivery frequency dominates JIT economics, and only the supplier-order uplift gives measurable protection against shortfall. In all, 650 packs (400 obsolete and 250 customer-rejected) require controlled disposition.

    2. Conclusions

      Inventory optimisation in an EV battery-pack plant is a cross-functional activity and not a Stores task. The answer is neither minimum nor maximum stock but the right stock for each material class, lead time, and consumption pattern. Local, reliable materials can be replenished frequently provided delivery frequency passes the c* test. Imported materials need lead-time-based protection. Shelf-life-sensitive items need FEFO and production-linked ordering monitored through A*. Obsolete and rejected packs need separate engineering and quality controls. The study met its four objectives at the level of a planning model, and the cost, floor-space, and regression results remain illustrative until validated with plant data.

    3. Suggestions

      For PPC, Stores, and Purchase: run a daily PPC Stores feedback loop so that the next material requirement reflects actual production; make FEFO, remaining shelf life, and A* mandatory inputs before purchase-order release; keep Cell Glue Part A and Part B balanced; and use controlled JIT for local, reliable materials while protecting long-lead imports. For Engineering and Quality: introduce an engineering-change inventory gate before SOP and maintain a customer-rejection RCA and closure dashboard with an owner and target date for each item. For Plant Management: review inventory, transport, and floor-space cost together, because lower stock is beneficial only when it does not create a larger shortage, transport, or line-stoppage cost, and sustain the controls with periodic 5S and expiry audits.

    4. Implications

      Plant management: controlled JIT and FEFO protect continuity while reducing blocked capital, floor occupancy, and expiry losses. PPC and Stores: actual-output feedback replaces plan-only ordering, so replenishment follows real consumption. Quality and Engineering: the gate and the RCA dashboard turn one-off clean-ups into preventive controls. Suppliers: early visibility of schedules and rejection feedback supports delivery and quality reliability.

    5. Future Scope

Future work can re-estimate the regressions on at least 90 days of field data, complete the shelf-life corridor with adhesive lead times, and add ABC XYZ classification, statistical safety stock, e- Kanban, supplier scorecards, and milk-run delivery optimisation once reliable history exists. Digital dashboards and predictive rejection analytics could replace manual tracking, and a study of battery recovery economics would value the recovered cells and components. A full financial model covering carrying cost, transport, shortage, and obsolescence

would quantify the return on the proposed controls, and post-implementation measurement over complete months would validate the results.

REFERENCES

  1. Chopra, S. (2019). Supply chain management: Strategy, planning, and operation (7th ed.). Pearson.

  2. Christopher, M. (2016). Logistics & supply chain management (5th ed.). Pearson.

  3. Christopher, M., & Peck, H. (2004). Building the resilient supply chain. The International Journal of Logistics Management, 15(2), 114

  4. Corley, K. G., & Gioia, D. A. (2011). Building theory about theory building: What constitutes a theoretical contribution? Academy of Management Review, 36(1), 12 32.

  5. Dhairiyasamy, R., Gabiriel, D., Bunpheng, W., & Chan, C.

    K. (2024). A comprehensive analysis of Indias electric vehicle battery supply chain: Barriers and solutions. Discover Sustainability, 5, Article 361.

    https://doi.org/10.1007/s43621-024-00595-7

  6. Efron, B. (1979). Bootstrap methods: Another look at the jackknife. The Annals of Statistics, 7(1), 126.

  7. Endurance Technologies Ltd. (n.d.). Annual reports and approved company information [Company documents used within confidentiality limits].

  8. Endurance Technologies Ltd., New Energy Plant. (2026). PPC daily reports, stores inventory snapshot (25 Aug 2026), and planning inputs [Unpublished internal records].

  9. Flyvbjerg, B. (2006). Five misunderstandings about case- study research. Qualitative Inquiry, 12(2), 219245.

  10. Ghare, P. M., & Schrader, G. F. (1963). A model for an exponentially decaying inventory. Journal of Industrial Engineering, 14(5), 238243.

  11. Goldratt, E. M., & Cox, J. (1984). The goal: A process of ongoing improvement. North River Press.

  12. Golhar, D. Y., & Stamm, C. L. (1991). The just-in-time philosophy: A literature review. International Journal of Production Research, 29(4), 657676. https://doi.org/10.1080/00207549108930094

  13. Green, S. B. (1991). How many subjects does it take to do a regression analysis? Multivariate Behavioral Research, 26(3), 499510.

  14. Gunasekaran, A., Goyal, S. K., Martikainen, T., & Yli-Olli,

    P. (1993). Modelling and analysis of just-in-time manufacturing systems. International Journal of Production Economics, 32(1), 2337. https://doi.org/10.1016/0925-5273(93)90005-6

  15. Heizer, J., Render, B., & Munson, C. (2017). Operations management: Sustainability and supply chain management (12th ed.). Pearson.

  16. India Brand Equity Foundation. (2026). Electric vehicle industry in India [Sector report].

  17. Künsch, H. R. (1989). The jackknife and the bootstrap for general stationary observations. The Annals of Statistics, 17(3), 12171241.

  18. Lee, H. L., Padmanabhan, V., & Whang, S. (1997). The bullwhip effect in supply chains. Sloan Management Review, 38(3), 93102.

  19. Little, J. D. C. (1961). A proof for the queuing formula: L = W. Operations Research, 9(3), 383387.

  20. MacKelprang, A. W., & Nair, A. (2010). Relationship between just-in-time manufacturing practices and performance: A meta-analytic investigation. Journal of Operations Management, 28(4), 283302. https://doi.org/10.1016/j.jom.2009.10.002

  21. Mahendran, S., Senthilkumar, A., & Jeyapaul, R. (2018). Analysis of lean manufacturing in an automobile industry a case study. International Journal of Enterprise Network Management, 9(2), 129142. https://doi.org/10.1504/IJENM.2018.093708

  22. Ministry of Environment, Forest and Climate Change. (2022). Battery Waste Management Rules, 2022. Government of India.

  23. Ministry of Heavy Industries. (n.d.). PM E-DRIVE and electric-mobility policy materials. Government of India.

  24. Nahmias, S. (1982). Perishable inventory theory: A review.

    Operations Research, 30(4), 680708.

  25. Ohno, T. (1988). Toyota production system: Beyond large- scale production. Productivity Press.

  26. Rubin, D. B. (1976). Inference and missing data.

    Biometrika, 63(3), 581592.

  27. Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J., Gatelli, D., Saisana, M., & Tarantola, S. (2008). Global sensitivity analysis: The primer. Wiley.

  28. Shah, R., & Ward, P. T. (2003). Lean manufacturing: Context, practice bundles, and performance. Journal of Operations Management, 21(2), 129149. https://doi.org/10.1016/S0272-6963(02)00108-0

  29. Silver, E. A., Pyke, D. F., & Thomas, D. J. (2017). Inventory and production management in supply chains (4th ed.). CRC Press.

  30. Slack, N., Brandon-Jones, A., & Burgess, N. (2022).

    Operations management (10th ed.). Pearson.

  31. Whetten, D. A. (1989). What constitutes a theoretical contribution? Academy of Management Review, 14(4), 490 495.

  32. White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48(4), 817838.

  33. Womack, J. P., & Jones, D. T. (1996). Lean thinking: Banish waste and create wealth in your corporation. Simon & Schuster.

  34. Wooldridge, J. M. (2020). Introductory econometrics: A modern approach (7th ed.). Cengage Learning.

  35. Xu, W., & Song, D.-P. (2022). Integrated optimisation for production capacity, raw material ordering and production planning under time and quantity uncertainties based on two case studies. Operational Research, 22(3), 23432371. https://doi.org/10.1007/s12351-020-00609-y

  36. Yang, M., Xiang, B.-R., Wu, X.-Y., & Liang, L. (2026).

    Battery production model decisions in the EV supply chain. Journal of the Operations Research Society of China. https://doi.org/10.1007/s40305-026-00689-1

  37. Yavuz, M., & Akçali, E. (2007). Production smoothing in just-in-time manufacturing systems: A review of the models and solution approaches. International Journal of Production Research, 45(16), 35793597. https://doi.org/10.1080/00207540701223410

  38. Yin, R. K. (2018). Case study research and applications: Design and methods (6th ed.). SAGE.

  39. Zou, J., Zhou, X., Wang, D., & Jin, S. (2025). Integrated optimization of battery manufacturing plant production and inventory capacities under market demand uncertainty. Manufacturing Letters, 44(Suppl.), 12191224. https://doi.org/10.1016/j.mfglet.2025.06.141